Exploratory Study of the Global Outcomes of the Older Americans Act Programs and Services Older Americans Act Programs and Services

Transcription

1 HEALTHCARE AND HUMAN SERVICES POLICY, RESEARCH, AND CONSULTING WITH REAL-WORLD PERSPECTIVE. Exploratory Study of the Global Outcomes of the Older Americans Act Programs and Services Final Report Prepared for: Submitted by: March 2013 Department of Health and Human Services, Administration on Aging The Lewin Group, Inc.

2 <Table of Contents Table of Contents Executive Summary... 1 Overview of Study Design Process... 1 Organization of Report... 1 Older Americans Act Title III Overview... 1 Table 1: OAA Title III Services... 1 Literature Review... 3 Methods and Overall Search Results... 3 Elements and Relationships to Consider... 4 Evaluation Design Options and Issues/Considerations... 5 Findings and Lessons Learned from Past HCBS Studies... 6 Available Data Sources and Data Elements... 7 Study Design Framework... 8 Logic Model... 8 Figure 1: Logic Model... 9 Research Questions... 9 Defining the Scope of the Study Proposed Study Design Figure 2: Proposed Study Design Image Proposed Study Design Description Limitations and Recommendations Limitations Recommendations Conclusion Appendix B: Study Framework and Design Appendix C: Limitations and Recommendations Executive Summary i

3 Executive Summary Executive Summary Overview of Study Design Process In the process of creating this study design, the research team identified a number of strengths and limitations to conducting a global outcomes evaluation of the Older Americans Act (OAA). This report proposes a study design of OAA programs. This study will assist AoA in identifying actions needed to better position the states for participation in a nationwide evaluation. Organization of Report This report is organized as an executive summary that references appendices that are the major sections of the project: Appendix A Literature Review; Appendix B Study Design; Appendix C Limitations and Recommendations. Older Americans Act Title III Overview Older American Act programs and services represent a significant federal investment in developing a comprehensive, coordinated, and cost-effective system of home and community-based services (HCBS) that enables adults to live independent and healthy lives in their homes and communities. The services under OAA Title III are described in the table below: Table 1: OAA Title III Services Title Title III-B: Supportive Services and Senior Center Programs i Services Title III-B funds a broad array of services that enable seniors to remain in their homes for as long as possible. These services include but are not limited to: Access services, such as transportation, case management, and information and assistance; In-home services, such as personal care, chore, and homemaker assistance; and Community services such as legal services, mental health services, and adult day care. This program also funds multi-purpose senior centers that coordinate and integrate services for older adults such as congregate meals, community education, health screening, exercise/health promotion programs and transportation. i Administration on Aging. Supportive Services and Senior Centers Programs. Retrieved from: Executive Summary 1

4 Executive Summary Title Services The purpose of the OAA Nutrition Program is to: Reduce hunger and food insecurity Promote socialization of older individuals Promote the health and well-being of older individuals and delay adverse health conditions through access to nutrition and other disease prevention and health promotion services. Title III-C: Nutrition Services ii Congregate Nutrition Services and Home-Delivered Nutrition Services provide meals and related nutrition services to older individuals in a variety of settings including congregate facilities such as senior centers; or by home-delivery to older individuals who are homebound due to illness, disability, or geographic isolation. Services are targeted to those in greatest social and economic need with particular attention to low-income individuals, minority individuals, those in rural communities, those with limited English proficiency and those at risk of institutional care. Nutrition Services Programs help older individuals to remain independent and in their communities. The OAA authorizes and provides appropriations to the Administration on Aging (AoA) for three different nutrition programs under Title III: Congregate Nutrition Services (Title III C1) Home-Delivered Nutrition Services (Title III C2) Nutrition Services Incentive Program (NSIP). Title III-D: Disease Prevention and Health Promotion Services iii Title III-E: National Family Caregiver Support Program iv Title III-D of the OAA provides grants to States and Territories based on their share of the population aged 60 and over for education and implementation activities that support healthy lifestyles and promote healthy behaviors. Health education reduces the need for more costly medical interventions. Priority is given to serving elders living in medically underserved areas of the State or who are of greatest economic need. The National Family Caregiver Support Program (NFCSP) provides grants to States and Territories, based on their share of the population aged 70 and over, to fund a range of supports that assist family and informal caregivers to care for their loved ones at home for as long as possible. The NFCSP offers a range of services to support family caregivers. Under this program, States shall provide five types of services: information to caregivers about available services, assistance to caregivers in gaining access to the services, ii Administration on Aging. Nutrition Services (Title III C). Retrieved from: iii Administration on Aging. Disease Prevention and Health Promotion Services (Title III D). Retrieved from: iv Administration on Aging. National Family Caregiver Support Program (Title III E). Retrieved from: Executive Summary 2

5 Executive Summary Title Services individual counseling, organization of support groups, and caregiver training, respite care, and supplemental services, on a limited basis These services work in conjunction with other State and Community-Based Services to provide a coordinated set of supports. Studies have shown that these services can reduce caregiver depression, anxiety, and stress and enable them to provide care longer, thereby avoiding or delaying the need for costly institutional care. Under this project, the Administration on Aging (AoA) seeks to study the impact of OAA programs and services, specifically services provided under Title III, on key outcomes, including HCBS use, health care use, community tenure, and long-term services and supports (LTSS) expenditures. OAA services and programs are diverse, often integrated and/or provided in combination with other services, funded through multiple funding streams, and administered and delivered by different state and locallevel agencies with varying data collection capacity. Therefore, studying OAA-funded programs and services is inherently challenging. Prior to developing this design, The Lewin Group (Lewin) conducted a literature review, developed and presented a study framework, and completed a study design. These steps are discussed below. Literature Review The development of a comprehensive OAA Global Outcomes Study Design required background research on prior studies about the impact of HCBS on the four outcomes of interest in this project: (1) HCBS Use; (2) Health Care Use; (3) Nursing Home (NH) Admission/Community Tenure; and (4) Cost Savings. To better understand potential study techniques and key variables related to these outcomes, Lewin conducted a review of the literature (Appendix A). Methods and Overall Search Results In order to conduct a thorough literature review, the Lewin team used articles from the Administration on Aging s (AoA) literature collection of studies compiled by S 3, previous literature reviews by the Lewin team, and a targeted updated search to identify new and additional relevant materials. After identifying approximately 900 articles, the Lewin team applied inclusion/exclusion criteria to identify: 19 key studies examining impacts of OAA services or new and useful findings on outcomes of interest related to OAA services, including HCBS use, health care use, Executive Summary 3

6 Executive Summary community tenure, NH admissions and costs used to inform the Findings and Lessons Learned Section; 100 additional studies and informational materials providing background information about OAA programs and services, key variables, and related outcomes; and, 21 key studies and papers from the program evaluation and social research literature. These studies, in combination with the 19 key studies, informed the methods for this design. Elements and Relationships to Consider LTSS presents a complex and fragmented delivery and financing system. A combination of federal, state and local sources fund a wide range of services to older persons residing in diverse communities and that require these services. To supplement these services, many older people rely on care provided by informal caregivers, caregivers paid out of pocket, or local agencies that provide services. Such a dynamic, multi-layered system requires tremendous coordination and constant adjustments to the changing policy environment and needs of the population. To study such a complex system requires understanding the challenges of defining constructs (e.g., funding, mix of services, unmet needs of individuals, etc.) and the means to measure these constructs (e.g., self-report, use of administrative records, standardized scales, the pros and cons of various proxy measures, and other considerations). Despite the inconsistencies in definition, we found a series of broadly defined variables that demonstrated statistical significance, or were previously found to have demonstrated significance in the studies background research, with the four key outcomes of interest to this review: 1) HCBS service use, 2) health care use, 3) community tenure/nh use, and 4) LTSS expenditures. We chose Andersen s Behavioral Model of Health Services Use (Andersen, 1995) to classify the variables into the domains listed below. This model has been widely used to determine variables tested in medical care studies (Andersen, 1995; Miller and Weissert, 2000). This model highlights predisposing, enabling, and need characteristics as variables that contribute to use of services, including HCBS, health care, and nursing home use. The literature review presented findings from numerous studies on how the characteristics, or variables, contribute to use of and/or cost of services. For each domain, we examined the impact the variables had on the aforementioned four key outcomes of interest. Predisposing Demographic variables including: Age, Sex/Gender, Race/Ethnicity, Education, Geographic Location, Living Alone, Marital Status, Having Children, and Spousal Use of Services Executive Summary 4

7 Executive Summary Enabling Need Demographic variables including: Income/Wealth, Home Ownership, Payment Method Informal Support: Availability and Use of Informal Support Caregiver Characteristics & Supports: Available Support for Caregiver, Caregiver Age, Caregiver Burden, Caregiver Awareness of Services Policy/Funding/Supply Level: State/Medicaid HCBS Funding, Proportion of LTSS Spending on HCBS, Number of Home Health Agencies, County Supply of NH Beds, NH Diversion/Pre-admission Screening Health characteristics variables including: Perceived Unmet Need, General Poor Health, Alzheimer's/Cognitive Impairment/Dementia, ADL Limitations, IADL Limitations, Comorbidities, MH Health Status Service Use, Mix and Spending Variables Acute: Hospital Admissions/Use, Emergency Department Use Post-Acute: Case Management, Timing of Care, NH Use (short-stay) LTSS: HCBS Use, Volume of Services Received, Combination/Bundle of Services, Use of State/Medicaid/Medicaid Wavier Services, Use of OAA III-B, Use of OAA III-E, Use of Title III-C, Timing of Care, Case Management, IADL Assistance, NH Use (Long-Stay) Lewin suggests that these variables be taken into consideration as part of the final study design. Study Design Options and Issues/Considerations Determining the outcomes attributable to a program or service presents a central challenge for any impact study. A study must not only use the best and most-efficient design, but it also must address the needs and values of persons served by the program and prove useful to informing policy and practice. Our review of previous HCBS studies and pertinent materials from the general evaluation literature identified several designs to consider in studying OAA programs and services: Randomized field experiments, considered the gold standard in research design for proving impact, offer a design most suited for new program models or service packages. This approach may be less appropriate for established and scaled up programs where no suitable control group exists or it might be unethical to withhold services from a control group. The literature indicates that randomized experiments are less informative when studying complex social interventions, such as OAA services and programs, with substantial variability in local applications. Executive Summary 5

8 Executive Summary Non-randomized comparison groups measure outcomes for two or more groups, but do not randomly assign participants to service receipt. Several HCBS studies have constructed comparison groups using different populations and analytical methods. Significant advances have been made in analytic methods that address vulnerabilities of comparison group designs and researchers have argued that results of this design can be extremely rigorous and emulate results of randomized experiments when certain criteria are met (Shadish & Cook, 2009). Various methodologies create comparison and treatment groups when using nonrandomized comparison designs, such as regression discontinuity and propensity score matching. These methods control for systematic differences in the comparison groups that would otherwise bias the results in a non-randomized design. Additionally, depending on the specifics of the research question and the context of the study, a variety of analytic techniques, such as survival analysis and structural equation modeling, can be applied to understand the impact and pathway of the outcome. The literature frequently recommends combining more than one analytic methodology. More than one strategy may be needed to answer all study research questions. Also, because all methods have limitations, synthesizing findings across multiple measures and methods may be needed to gain a full understanding of the program. In addition to the complex challenges associated with designing a valid and reliable impact study, we must consider the dynamic nature of the LTSS delivery system. At times programs expand based on new policy guidance, while, at other times, they contract in response to budgetary constraints. Further, each State and locality has different funding systems, policies, practices, and client needs that must be accounted for in any design. Lewin compiled a list of key questions to ask states in order to develop a study design that could best take into account the variation within states. This list is included in the Study Design (Appendix B). Findings and Lessons Learned from Past HCBS Studies Our synthesis of previous HCBS studies suggests several key findings, lessons learned, and recommendations to consider in studying the impact of OAA and non-oaa HCBS programs. These findings suggest that designing a study that produces credible, solid evidence of results will be of critical importance to AoA, the Aging Network, and service recipients and families who rely on OAA and non-oaa programs and services. Increasingly, stakeholders seeking continued or expanded funding for health and social programs must be able to make the business case for the investment, with rigorous research demonstrating success in improving peoples lives and achieving more effective/appropriate use of economic resources. In addition to providing evidence of an impact, a study design that allows for exploration into the pathways through which Executive Summary 6

9 Executive Summary the impact is made (e.g., case management, targeting of services, individual belief about HCBS, etc.) will allow policy leaders to make informed decisions about how to allocate future funding in a way that results in the greatest possible impact on outcomes. The primary findings are listed below: OAA and non-oaa services work better when offered as part of a comprehensive system to meet individual needs. Coordinated systems of care at the national, state, and agency level are important considerations when evaluating the effects of OAA and non-oaa social support activities; New studies highlight the effectiveness of moderate services, including OAA home-delivered meals and IADL help; Availability of services and resources in the community play an important role in affecting OAA and non-oaa HCBS outcomes. For OAA and non-oaa services, the greatest HCBS impacts have been on community tenure, rather than health care utilization. Community tenure is associated with family caregiver availability and individual characteristics of both the caregiver and the consumer. An assessment of HCBS outcomes must consider the effect of OAA services, controlling for the impact of non-oaa services. Service use is affected by many individual characteristics, some of which are difficult to measure (e.g., changing needs over time). Virtually every study concluded that further research is needed to develop understanding of the complexities inherent in the LTSS delivery system. Available Data Sources and Data Elements Our review identified numerous data sources that have been used in recent related HCBS studies, the types of data available from each source, and examples of studies using the data. These include OAA-specific, state, national and qualitative data sources. In addition, we provide examples of surveys of HCBS programs that could potentially be used to capture some types of program information that have been identified in the literature as important to consider, but are not included in existing data sets. A high-quality study design hinges on valid and reliable data. Regardless of how well a study is designed, the outcome is only as good as the data used. While available data sources provide a cost-effective and efficient way for collecting data, all sources reviewed were found to have limitations in terms of the measures/indicators collected, the quality of the data, the frequency of the collection, or comparability of the measures/indicators over localities and states. Many studies have used supplemental surveys, questionnaires or interviews to collect additional data that complement the Executive Summary 7

10 Executive Summary data collected through regular reporting systems to allow for more robust evaluation results. While supplemental data collection methods may be costly, they are likely the only way to ensure the inclusion of high quality data on certain key variables (e.g., availability of informal caregivers, level of unmet need, functional status, etc.). Study Design Framework The Study Framework and Design (Appendix B) describes Lewin s proposed approach for understanding the impact of OAA programs and services on the aforementioned outcomes of interest. Building off of the literature review, this document includes a discussion of the challenges and trade-offs that must be considered in the selection of variables, research questions, study design options and secondary data sources, in order for AoA to endorse a rigorous impact study of OAA programs and services. This design focuses on the impact of services provided through OAA-funded HCBS programs authorized under Titles III-B, C, D, and E of the OAA. These programs include a range of supportive services, nutrition services, health promotion and disease prevention programs, as well as services for family caregivers. While the mix and type of services offered differ by state and locality, the vast majority of OAA funding is used for the provision of nutritional services. Logic Model To best understand how OAA program activities relate to short and long-term outcomes, a logic model was developed in collaboration with ACL staff (Figure 1). The logic model depicts the inputs, activities, outputs, and outcomes of OAA-funded HCBS programs and services. We have included non-oaa funded activities (e.g., informal services, state-funded HCBS, private pay) as these activities often occur in combination with OAA-funded activities and impact the key outcomes of interest for this project. Executive Summary 8

11 Executive Summary Figure 1: Logic Model Research Questions Based on the expected outcomes depicted in the logic model, we propose the following research questions. What is the impact of OAA-funded HCBS programs and services on: Community tenure Health care utilization Costs of care for older adults (e.g., LTSS, health care costs) Physical, mental, and emotional health and wellness (e.g., preventive measures) of care recipients and caregivers Unmet needs among older adults Caregivers (e.g., strain, burden, depression, health, etc.) Coordination of services (e.g., care management) Executive Summary 9

12 Executive Summary In addition to answering questions about the impact of service use on desired outcomes, we recommend that AoA consider a design that measures the strength of association of covariates/intervening variables on the key outcomes including, but not limited to, mix, type, and intensity of services. This step would answer the following sample questions: What is the impact of OAA services alone or in combination with services paid for by other sources? What is the impact of service mix and intensity on outcomes of interest? [if possible, we will isolate OAA services] What subgroups had the most favorable outcomes? (e.g., health conditions, demographics, functional status) Defining the Scope of the Study A critical methodological decision hinges on how non-oaa funded services are factored into the study design. For example, does the design consider the OAA service package alone or in combination with a similar service package that is funded by alternative sources? This decision will impact the comparison group, sample size, research questions and generalizability of the findings. Another important issue discussed frequently in the literature, and one that we are confronted with in this design, is the level of exposure to HCBS services (i.e., mix, duration, intensity and timing). Consideration of this step will ensure that individual variations are factored into the final models. Several studies that were reviewed developed exposure algorithms that may be applicable for use in the final models to address this issue (NYSOA, 2010). Furthermore, an important issue raised in the literature is the impact of certain policy or other contextual factors on individual outcomes (e.g., nursing home admission or HCBS use). These measures include the amount of state spending on Medicaid HCBS, the number of nursing home beds in a geographic area, availability of Waiver slots, the number of home health agencies in a particular area, and the use of pre-admission screening prior to NH use. Several of the key variables proposed for the OAA study are well established measures (e.g., sex as male/female/unknown). However, the literature points to several limitations of many of the measures (e.g., informal care measures, ADL/IADL limitation measures), which must be considered prior to including them in the study. Executive Summary 10

13 Executive Summary Proposed Study Design We suggest that the design approach with the greatest potential for delivering conclusive and actionable results would be a quasi-experimental design that includes: 1) a retrospective study component drawing on existing individual data from federal sources; and 2) a prospective study component using data to be collected during the study period. Evaluators of HCBS and other health and social service programs usually rely on quasi-experimental techniques as the best impact study design and most powerful alternative to random experiments. We know from the literature that demonstrating impacts on HCBS expenditures can take several years, with a lag between increased spending on HCBS expansion and savings from this investment (Kaye, LaPlante, & Harrington, 2009). Thus, we recommend a study period of seven years, a plausible option given the retrospective/prospective design option. This study period time can be modified depending upon time, data, or funding constraints. To reduce bias and error in assessing program outcomes, we recommend incorporating multiple methods, measures, and data sources. In using a quasi-experimental design with a matched comparison group, a suggested method includes propensity score matching for constructing a matched comparison group in the final design. This will allow comparisons of older adults who had an outcome of interest with older adults who are matched on certain key characteristics relevant to the outcome, to determine whether OAA services contributed to the difference in outcome. The matching process will allow us to control for factors that are believed to contribute to different outcomes (e.g., gender, diagnosis, multiple chronic conditions, etc.). The key to unbiased inference in this approach, as in any quasi-experimental design, is proper specification: choosing a set of key factors to be controlled so that uncontrolled variables are approximately random, i.e., not correlated with the controlled variables or the treatment (Achen, 1986). Knowledge about the factors that affect receipt of services and factors correlated with outcomes (i.e., community living tenure and healthcare utilization) identified in the literature review are essential to the success of the design. The research team proposes using this design in a study of 3 states. A primary purpose of the study will be to identify a core set of variables that all states should collect. States have been granted great flexibility regarding program administration and data collection. As a result, states lack uniform and comprehensive data. Conducting this study, as discussed in the Recommendations section, can address that issue. Executive Summary 11

14 Executive Summary Figure 2: Proposed Study Design Image Executive Summary 12

15 Executive Summary Proposed Study Design Description The above graphic depicts the proposed design for the Study of the Global Outcomes of Older American Act Programs and Services. The goal is to measure the impact/association of OAA service use by older adults on four key outcomes (health care utilization, HCBS expenditures, NH admissions and community tenure). Further, the design allows for the examination of various service use trajectories over time (e.g., HCBS user in a NH, NH user exits NH, etc.) and the extent to which OAA programs and services are associated with these trajectories. The details of the study design are discussed below. Intervention: Receipt of OAA services, alone or in combination with other HCBS. Target Population: The target population for this study is older adults age 60 and above (those eligible for OAA services). In the selection of the cohorts in stage 1, the sample will be selected and then connected back to those who received OAA services in stage 2 of the design. Those who have received OAA services are the treatment, or intervention, group and those who are not matched to OAA service use are the comparison group. Pre-Stage: Three states will be selected through an RFP process to participate in this study design. The states will have the data capacity and availability to participate in the stages outlined below. Stage 1: The proposed design takes place over a six-year study period and consists of three stages. In Stage 1, (depicted in the blue box in the middle), the study sample is selected using MDS, Medicaid/Medicare claims data, and OAA or state-funded services data, that cover a one-year period of time ( ) and an entire state (e.g., Georgia). Three cohorts of older adults will be identified from this dataset: Cohort 1 will be a sample of older adults who have been admitted to nursing homes; Cohort 2 will be a sample of older adults enrolled in any HCBS [Medicaid, state-funded, or OAA]; and, Cohort 3 will be a group of individuals who were not admitted to a nursing home and did not receive HCBS. This group will be matched to older adults in Cohorts 1 and 2 on certain key variables to ensure comparability (propensity score matching). Cohort 3 will serve as the comparison group. Stage 2: During Stage 2 of the study (represented in the purple box on the left), the HCBS service use by Cohort will be measured. This will be accomplished through the analysis of Medicare/Medicaid Claims, POMP, State OAA administrative records, HRS and any other available sources. The study period covers Data will be analyzed by service type, mix, and intensity of service use. A group of individuals will be non-users. Stage 3: In the final stage of the study, Stage 3, (illustrated in the green box on the right), using the same combination of data sources, three additional years of service use Executive Summary 13

16 Executive Summary and client disposition status post-intervention will be examined. In sum, the study will have a longitudinal data base ( or the most current year available to us) that includes three unique cohorts of individuals: nursing home users at baseline, HCBS users at baseline, and non-users at baseline. This combination will allow for both within-group and between-group data modeling. Results will demonstrate if receipt of OAA services, whether alone or in combination with other services, will directly impact: health care utilization, HCBS expenditures, nursing home (NH) admissions and community tenure. Limitations and Recommendations The Administration on Aging (AoA) and The Lewin Group (Lewin) hosted an Expert Group meeting, conducted several key informant interviews with additional stakeholders and held calls with states about their Older Americans Act (OAA) data systems. This feedback yielded a better understanding of the limitations of this study design and the recommendations to address them (Appendix C). Limitations The following major limitations were identified in those discussions: State Older Americans Act (OAA) Data: A select group of states (FL, GA, MA, MN, and OH) were interviewed about their OAA data collection policies and procedures. These interviews highlighted some of the limitations of state OAA data, along with identifying states with more comprehensive data systems from our small sample. Specific limitations that impact the study design included: The absence of Social Security Numbers (SSNs), which provide the most reliable method to match to Medicare and Medicaid claims. Lack of information regarding the reasons for termination of services, which limits the ability to measure community tenure as an outcome. Informal Caregiver Data: As discussed in the expert group meeting and key informant interviews, difficulties will be encountered obtaining caregiver information, due largely to the non-existence of such data. Private Pay Data: This is a limitation for both the treatment and comparison groups. States do not uniformly collect information on private pay services received by OAA users. Therefore, this information would not be readily available for the proposed treatment or comparison groups. Recommendations While this study can be conducted, it will require complicated data gathering, cleaning and merging to overcome many of the data limitations. It will also be heavily Executive Summary 14

17 Executive Summary dependent on the strength of the chosen states OAA data system and timeline when the states migrated to a statewide data system. Based upon the limitations of data sources available, and feedback from expert group participants and key informants, the recommended approach is outlined below: Step 1: Issue RFP to conduct study AoA issues an RFP to carry out the proposed study. The awarded research team would work with AoA to prepare for and conduct the study. Step 2: Convene Technical Working Group Within three months of the contract begin date, AoA in collaboration with the awarded research team establishes a Technical Working Group (TWG). A TWG can assist the research team in identifying the minimum key variables necessary for a retrospective/prospective design. Step 3: Issue RFP to states Prior to the completion of step 2, AoA issues an RFP identifying up to three states willing to participate in organizing existing data at the state level. Proposed criteria for this RFP are outlined in the Limitations and Recommendations document. Step 4: Assist states / Collect available secondary data Step 4 has two concurrent components that will begin upon awarding three states grants to conduct the study: (1) AoA and the research team will work with the three selected states in gathering and preparing their OAA administrative data; and (2) while the selected States are preparing and assisting in the collection of data, the research team should assist the states as necessary. In addition, the research team will begin preparations for linking the state-level data with other national datasets. Step 5: Expand work to include primary data collection Step 5 includes expanding the work described in steps 2-4 in the selected states to include primary data collection, as funds and time allow. Primary data collection will strengthen the data by gathering detailed information on such key aspects of caregiving as informal supports. This step could be accomplished through either a state-wide random selection of OAA participants, or targeted data collection that occurs with select AAAs in the chosen states. Step 6: Design and field the study The research team selected in Step 1 should work with AoA and the three participating states in finalizing the study design and fielding the study with data collected at the state and federal level. The proposed design in the Study Framework and Design should be the foundation for this design, and modified as appropriate given available state secondary data and any additional primary data. Executive Summary 15

18 Executive Summary Step 7: Reconvene TWG and present findings Once the exploratory study is conducted in the three selected states reconvene the TWG and present results for feedback and discussion. TWG members, in collaboration with AoA and the research team, will define a core set of variables that AAAs will employ to ensure consistent data collection across agencies. Step 8: Training & technical assistance for data collection With changes to SPR data requirements, many states will require intensive T&TA as they make updates to their current data systems. However, some states will be more prepared than others or already collect the core variable requirements, while others will require a complete overhaul. Conclusion Conducting this study can position AoA to design and implement a more comprehensive evaluation of OAA services. As all states begin collecting newly required data elements, AoA might consider how this new data can be used in a national evaluation, or an evaluation conducted in a random selection of states. This evaluation could result in definitive conclusions providing a full picture of the impacts of OAA Title III services to date. Specifically, AoA will better understand the impact of OAA services across the board in promoting positive outcomes for service recipients, including increased community tenure and decreased health care use. This evaluation will also examine the potential cost savings of OAA service use. Additionally, it will enhance capacity for ongoing research and provide information about the programs that AoA and the Aging Network can use in future program planning. Executive Summary 16

19 Table of Contents Study Design Options Findings from Early HCBS Evaluations Theoretical Considerations Findings and Lessons Learned from Past HCBS Studies Conclusion Appendix A-A: Available Data Sources and Data Elements Appendix A-B: Detailed Key Variable Table Appendix A-C: Detailed Bibliography Tables Table 1: Data Sets with OAA/Aging Network Specific Data Table 2: Other State Specific Data Sets Table 3: National Data Sets and Other Sources i

20 Introduction This report presents The Lewin Group s findings from the review of materials (Subtask 2.1 of the project to design an Exploratory Study of the Global Outcomes of Older Americans Act Programs and Services). In this overview, we describe our methods and present findings from our review of materials. These include findings related to: (a) a study design that may be suitable for assessing OAA program impacts, taking into account the complexity in measuring impact; (b) findings and lessons learned from past HCBS studies; (c) elements and relationships to consider in developing a program model and planning a study of HCBS impacts (including HCBS goals, policies, strategies, impacts, and research questions); and (d) available data sources identified for us in examination of these issues (Appendix A-A). Methods and Overall Search Results We found relevant studies, reports, and other published materials through a variety of methods. There were two phases to our search: 1. We utilized the collection of articles previously compiled for AoA by S 3 and other materials provided by AoA. We augmented these study reviews with a targeted search for new and additional studies (published and unpublished) on the impacts of HCBS on community tenure, health care utilization, and/or economic/systems outcomes, through: a. A search on PubMED and Google Scholar using key search terms related to this literature review, and b. Browsing research articles and abstracts in the Clearinghouse for Home and Community Based Services ( 2. Additional studies were obtained through other sources (e.g., lists, reviewing sources cited in reviewed studies, other studies known to the research team, and recommendations from the expert group, key informants, and state OAA representatives). From this, we scanned approximately 900 articles/studies and narrowed the review to the studies below. Inclusion criteria included: Materials providing background information about OAA programs and services. General review of HCBS studies that included extracting relevant information to inform this report: previous evaluations of OAA services and studies of HCBS use on community tenure, health care utilization, and/or economic outcomes were used. These studies were most directly relevant to this project. 17

21 In-depth review of studies where we extracted information on the data sources, detailed findings for all outcome measures, and the authors recommendations for future research. Two researchers read the full text of the studies and reviewed each entry. For evaluation methodologies, we paid particular attention to 19 studies that specifically examined impacts of OAA services or that provided new and useful findings to help inform a study design. The in-depth reviewed studies included: Studies that specifically examined OAA programs and services, including research conducted as part of the Administration on Aging s Performance Outcomes Measurement Project (POMP), and related studies by the same authors that specifically examined outcomes of OAA services, along with other HCBS, in numerous states (e.g., Florida, New York, Georgia, and Rhode Island). Findings from the Community Tenure study, which examined receipt of OAA services, along with other HCBS, and has been nationally recognized for its methods. This 5-year prospective study tracked older adults in Kansas who had applied for nursing facility admission and received a pre-screening admission assessment under the State s Client Assessment Referral and Evaluation (CARE) Program. The study tracked CARE participants community tenure and outcomes when community tenure ended (Chapin et al., 2009). Select studies of non-oaa HCBS that could be particularly useful in developing a study design. These were recent studies that: (a) measured the impacts of HCBS on community tenure and/or health care utilization; (b) used rigorous methods and large samples so that the results are reasonably generalizable; (c) examined a complex set of factors to rule out alternative explanations for program results; (d) examined relationships or factors not previously examined in the literature. Over 100 articles presenting background research, prior HCBS studies and evaluations, and research on factors predicting HCBS use, health care use, and nursing home utilization. These include meta-analyses that examined predictors of nursing home admission (Gaugler et al., 2007) or other adverse outcomes (Miller & Weissert, 2000). Other studies examined targeted groups experiences with the outcomes of interest, such as people with dementia, older adults in rural areas, and African American women. 21 relevant materials from the general program evaluation/social research literature. These materials were used to enhance understanding of evaluation design options suggested by the HCBS studies and to identify other potential design options that may be appropriate to consider. These evaluation design-related materials included: evaluation literature known to Lewin and AoA, materials recommended in OMB guides, and additional items found through targeted searches for materials about specific methodologies that appeared most relevant. 18

22 There were many studies that, while informative, did not meet our inclusion criteria for further review. These include: Studies of particular program strategies or interventions (e.g., adult day care, assistive technology, care management, participant direction, food assistance programs); Studies related to the direct service workforce, studies of informal/family caregivers; Overview articles/issue briefs; State specific program overviews; Research on demographic trends and predictors of health outcomes; Studies related to needs assessments; Studies related to medical, exercise or therapeutic home-based interventions; Studies not written in English and/or not conducted in the United States We did not apply exclusion criteria based on the time in which the study was conducted or published to capture the early HCBS findings from the Channeling Demonstration. Study Design Options In this section, we summarize our findings from the reviewed HCBS studies and program evaluation literature on possible study design options for AoA to evaluate the impact(s) of the Older Americans Act (OAA) and services on older adults and caregivers. Additionally, we must account for the combination of OAA programs and services with non-oaa HCBS programs and services. Strategies in Evaluating Aging and Disability Programs A study that seeks to evaluate the effectiveness of a social program, frequently referred to as a program evaluation, is different from other types of research (Krause, 1996; Rossi, Freeman, & Lipsy, 1999). In what is sometimes described as pure research, social scientists start with hypotheses, gather data, and then try to form generalizations about the results. A search for theory is the primary function, and this quest for theory requires the use of research procedures in a fashion that often differs from what generally occurs within program evaluations. For example, correlations that might produce a scientific journal article could easily be meaningless to an agency director who is trying to decide whether to spend additional dollars on a particular program. While the major focus of studies evaluating program outcomes is to determine whether outcomes can be attributed to the program (GAO 1991; GAO 2012), they are also 19

23 concerned with identifying any unintended outcomes and understanding how the program works with other programs (Treasury Board of Canada, 1998). DIFFERENCE BETWEEN PROGRAM EVALUATION AND OTHER RESEARCH Program evaluation is the use of social research procedures to systematically investigate the effectiveness of social intervention programs. Rossi, Freeman, & Lipsey, 1999 Program evaluation, or evaluation research, refers to the research procedures and techniques used to examine the effectiveness of social programs. Program evaluation, in other words, is a process that generates the information used to describe what a program is doing and how well it does it. Krause, 1996 GAO (2012) recommended that five key steps, similar to those presented by Krause (1996), should be taken into consideration before data are collected in a program evaluation: 1. Clarify understanding of the program s goals and strategy. 2. Develop relevant and useful evaluation questions. 3. Select an appropriate evaluation approach or design for each evaluation question. 4. Identify data sources and collection procedures to obtain relevant, credible information. 5. Develop plans to analyze the data in ways that allow valid conclusions to be drawn from the evaluation questions. Each of these steps is crucial and should not be overlooked. As a tool for understanding how program activities relate to short-term and long-term outcomes, evaluation materials typically recommend using a program logic model (GAO, 2012). A review of materials, along with discussions with program stakeholders, is essential to gathering the information needed to develop a successful program logic model and evaluation design. Content analysis or a summary of the content of existing materials, such as that undertaken in this report, has several applications in program evaluation (GAO, 1989). These include identifying program goals, describing program activities, and determining program results. 20

24 Research Designs Research designs fall into several groups discussed in depth in this section: randomized; non-randomized comparison group; time series/longitudinal; and alternate causal explanation elimination. A frequent recommendation in the evaluation literature is to combine two or more research designs. GAO (2009) recommended, Since all evaluation methods have limitations, our confidence in concluding that an intervention is strengthened when the conclusion is supported by multiple forms of evidence (pg ). Some of the HCBS studies cited used more than one type of design; for example, a survival analysis study was used with a matched comparison group in the Kansas Community Tenure Study (Chapin et al., 2002). The Treasury Board of Canada (1998) advised that using more than one evaluation strategy is desirable to increase support for inferences about program impact. This is because, Generally speaking, no single evaluation strategy is likely to yield enough evidence to answer unambiguously the questions posed by the evaluation (Treasury Board of Canada 1998, p. 28). Because an evaluation study generally addresses several issues at a time, more than one strategy may be needed to answer all research questions (Treasury Board of Canada, 1998). Trochim (2006) noted that the currently prevailing philosophy in science recognizes that all observation is fallible and has error and that all theory is revisable. This view emphasizes the importance of multiple measures and observations, each of which may possess different types of error and the need for triangulation (i.e., synthesizing findings) across these multiple error-containing sources to try to gain a full understanding of the program. Shadish and Cook (2009) described pattern matching designs, which counter the unfortunate notion that researchers should choose from a small and fixed set of designs (pg. 623). While researchers often focus on minimizing threats to validity, pattern matching designs attend to a less often noticed piece of advice, to predict a diverse pattern of results whose strong testing might require multiple nonrandomized designs, each with different presumed biases (pg. 623). Shadish and Cook (2009) quoted Campbell & Stanley s 1963 advice that the more numerous and independent the ways in which the experimental effect is demonstrated, the less numerous and less plausible any singular rival invalidating hypothesis becomes (pg. 623). They recommended that when an ideally implemented experiment or quasi-experiment is not feasible, researchers assemble more than one design that predicts a pattern of causal results, because fewer alternative explanations are plausible if the results match the predicted pattern. An example discussed by Shadish and Cook (2009) was measuring the same outcome via a randomized and nonrandomized comparison group design then comparing results. They referred to Cook et al. (2008), who suggested a stringent set of seven 21

25 criteria for good comparisons of results obtained from a randomly formed and a nonrandomly formed control group. 1. The studies compare a randomly formed control group and a non-randomly formed control group. 2. The randomized and nonrandomized experiment both estimate the same estimator (e.g., average effect of treatment on the treated or intent to treat). 3. The randomized and nonrandomized groups should differ from each other only in method of assignment. 4. The person estimating results from the nonrandomized study should not know the results from the randomized experiment. 5. The randomized experiment should be an exemplar of its kind, not subject to large attrition or partial treatment problems. 6. The quasi-experimental design should similarly be an exemplar of its kind, without attrition or partial treatment problems, with focal local controls and good pre-test measurement of variables related to treatment and outcome. 7. A defensible standard for what counts as a match in randomized and nonrandomized results is used. This is difficult both because reasonable people might disagree on substantive criteria that would make a difference to policy decisions and because statistical criteria will inevitably be subject to power problems (pg. 622). Cook and colleagues (2008) showed that when most or all of the seven criteria were met, results from different kinds of nonrandomized experiments matched results from randomized experiments. This was true for regression discontinuity designs, welldesigned nonrandomized experiments with focal local controls and stable matching, and statistical analyses such as propensity scores. Randomized Field Experiment Many HCBS studies, using randomized assignment to test the effectiveness of a particular program, or of HCBS overall, have encountered serious implementation problems that lead to inconclusive results. The early HCBS studies revealed many limitations to the use of randomized experiments in the study of HCBS programs. These include: lack of clarity over whether the actual experimental intervention was likely to have its intended effect, small estimates of results due to similarities in services received 22